ISPRS Journal of Photogrammetry and Remote Sensing 74 (2012) 153–162 Contents lists available at SciVerse ScienceDirect ISPRS Journal of Photogrammetry and Remote Sensing journal homepage: www.elsevier.com/locate/isprsjprs Temporal characteristics of thermal satellite images for urban heat stress and heat island mapping ⇑ Janet E. Nichol , To Pui Hang Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University, Kowloon, Hong Kong article info abstract Article history: The reconstruction of urban climate is still challenging to climatologists in spite of over five decades of Received 6 August 2011 research including direct data measurement and model building. Methods for measuring and monitoring Received in revised form 21 April 2012 urban climate have strengths and weaknesses depending on the application. The mapping of patterns of Accepted 5 September 2012 urban heat stress over a city is not useful if the patterns depicted apply only to the time of data acquisi- Available online 5 November 2012 tion. Since thermal satellite sensors can now provide detailed temperature data covering whole cities and beyond, their adoption in urban planning depends on demonstrating their relevance to commonly pre- Keywords: vailing conditions. This research investigates and presents a methodology based on four summertime Thermal ASTER thermal satellite images of Hong Kong for urban heat stress mapping at detailed level. It demon- Satellite Urban strates that satellite images obtained under certain climatic conditions, and accompanied by adequate Climate ‘in situ’ ground data, can provide a basis for an operational heat stress mapping system. The temporal limitation of thermal satellite images is examined for both day and nighttime images by comparison of image-derived air temperatures with ground data representing extended periods and other hot days and nights outside the image acquisition times. The nighttime images were found to be more represen- tative of air temperature at other times than the daytime images, due to a more stable boundary layer, with lower wind speeds and temperature inversion at night. The nighttime images showed high and sig- nificant correlations with ground level air temperatures for an average 13-h period surrounding the image time 10.42pm, from 6 pm to 4–8 am the next day. Additionally they were highly and significantly correlated with ground air temperature distributions on 93% of all hot summer nights in the same years. Therefore the nighttime images can be considered representative of a commonly occurring summer nighttime situation in Hong Kong, and can be used to determine the locations of areas where tempera- tures commonly exceed hot weather warning thresholds. Notably, the images were better able than cli- mate stations to represent areas in the urbanized Kowloon Peninsula and several smaller satellite towns which exceeded hot weather warning thresholds. Many areas exceeded the thresholds, even when no hot weather warning was in force, due to the unrepresentative location of climate stations. The images were also more able than climate stations to indicate the hottest and coolest areas over the Hong Kong terri- tory, thereby enabling measurement of the magnitude and extent of the urban heat island. Ó 2012 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS) Published by Elsevier B.V. All rights reserved. 1. Introduction ically reach 33 °C on the hottest summer days, cooling to ca. 26 °C in urban, and 24 °C in rural areas at night, but little research on 1.1. Hong Kong’s urban heat island population heat stress has been done in tropical cities. Dousset et al. (2011) describe the 2003 European heat wave with daytime Rising global temperatures bring into question the sustainabil- maxima and nighttime minima exceeding 37 °C and 25 °C respec- ity of tropical and sub-tropical cities (McCarthy et al., 2010), espe- tively over 9 consecutive days, which resulted in 4867 excess mor- cially those such as Hong Kong where dense and high rise buildings talities in Paris, and 70,000 over the whole of Europe but mortality are accompanied by an intense urban heat island (UHI) effect (Fung appeared to be more closely related to nighttime than daytime et al., 2009; Nichol et al., 2009). At 22°N Hong Kong lies at the temperatures (Robine et al., 2008). In Hong Kong, Kowloon, with boundary between the tropics and sub-tropics, experiencing very a population of over 2 million is the largest continuous urban area hot humid summers and warm dry winters. Air temperatures typ- and is separated from the rural New Territories by mountain ranges. Parts of Kowloon contain the highest urban population ⇑ Corresponding author. Tel.: +852 2766 5952; fax: +852 2330 2994. densities in the world, with over 60,000 persons per sq km (Ng, E-mail address: [email protected] (J.E. Nichol). 2009). Reduced ventilation, high temperatures and the blocking 0924-2716/$ - see front matter Ó 2012 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS) Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.isprsjprs.2012.09.007 154 J.E. Nichol, P.H. To / ISPRS Journal of Photogrammetry and Remote Sensing 74 (2012) 153–162 of sea breezes by tall buildings on newly reclaimed land, the so vance to times and dates other than the immediate imaging time, called ‘wall effect’ (Wong et al., 2011) are contentious issues. Plan- especially in cloudy regions where good quality thermal images ners urgently require empirical and spatially comprehensive cli- may be obtained only infrequently. In summary, the satellite-de- matic data to support structure plans and policies but traditional rived surface temperature may be able to represent air tempera- methods of UHI analysis, the use of fixed stations and/or vehicle ture, and its relevance is dependent on the establishment of a traverse have been unable to provide such data because the data robust relationship between the surface and adjacent air at the im- collected are spatially incomplete. age time. If this relationship can be demonstrated to be temporally stable, and representative of other times and dates where com- 1.2. Data sources for urban climatic analysis monly occurring climatic conditions occur, it is suggested that the image can be used as a working model of the spatial distribu- Data sources for measuring and estimating urban climate in- tion of air temperatures over a city for planning purposes. clude automatic weather stations (AWSs), vehicle traverse, climate simulation models, and satellite observations. However in Hong 1.3. Objectives Kong, as in many other cities, there are no AWS in core city areas, since the Hong Kong Observatory climate station, historically re- This research tests the relevance of temperature patterns ob- garded as the most ‘urban’ is situated in park-like grounds where served on thermal satellite images both at the image time, and at temperatures are consistently lower than in nearby densely built other times based on Stoll and Brazel’s (1992) assertion that corre- areas where people live and work. Furthermore, the climatic data lations between surface and air temperatures are dependent on used for predicting the effects of temperature on mortality in Hong atmospheric mixing, mean wind velocity and the thermal proper- Kong e.g. number of deaths when the Net Effective Temperature ties of surface materials. Thus, on imagery obtained during a stable (NET) exceeds 26 °C, are based on temperatures at the (unrepre- atmosphere and low wind speeds (e.g. at night, when turbulence is sentative) HKO station. Vehicle traverse data are similarly incom- greatly reduced (Kawashima et al., 2000; Voogt and Oke, 2003)), plete spatially, and suffer from a time lag during the data and surface effects minimized, reduced advection would result in collection. Climate simulation models suffer from the complexity a stronger relationship between the surface and the adjacent air. introduced by the roughness of the urban canopy layer structure, Thus the image may provide a representation of air temperatures the difficulty of mapping surface materials in detail over a city accurate enough for a working model of a city’s urban heat island (Herold et al., 2006) and often lack anthropogenic heat contribu- situation. For example Fung et al. (2009) observed an R2 of 0.78 be- tions (e.g. Ashie, 2007). Additionally most urban climate models tween over 14,000 air temperature points from a mobile vehicle represent local (i.e. urban canyon) scale processes (Mills, 1993; traverse, and a night-time ASTER thermal image, when ground Arnfield and Grimmond, 1998; Arnfield, 2000), or neighborhood measured air temperature points falling within the same pixel scale processes (Grimmond et al., 2010) which are difficult to scale were averaged. The temporal deficiency of satellite images, being up to land use zones or whole cities. Thus they do not easily predict acquired at a single point in time is also likely to be minimized un- temperature patterns at city scale, or permit studies of urban heat der such conditions, when heat island patterns depicted may per- islands. Very recent developments in CFD modeling are able to sist for several hours, although this has not been tested previously. simulate energy balance and air temperatures across a whole city, It is relevant because UHI patterns present under a stable atmo- but use very large databases and are highly computer intensive sphere may represent a typical, or baseline heat island situation (Ashie and Kono, 2011). Other recent developments include the for a city i.e. a heat island distribution pattern which is the most downscaling of regional climate models such as WRF to city and commonly prevalent. district scale (Chen et al., 2011), but they incorporate limited land In light of the urgent need for information in Hong Kong dis- cover types e.g. the WRF urban modeling system incorporates only cussed above, this study uses a daytime and a nighttime ASTER three land cover types.